NOTE: This is a draft in progress, so that I can get some feedback from early reviewers. It is not yet ready for learning.
Bitcoin Scripts can go far beyond the relatively simple financial instruments detailed to date. They're also the foundation of most complex usages of the Bitcoin network, as demonstrated by these real-world examples of off-chain functionality, drawn from the Lightning Network examples in BIP 112.
The Lightning Network is a payment channel that allows users to take funds off-chain and engage in numerous microtransactions before finalizing the payment channel and bringing the funds back into Bitcoin. Benefits include lower fees and faster transaction speeds.
BIP 112 contains a few examples of how these off-chain transactions could be generated, using Bitcoin locking scripts.
The trick with Lightning is the fact that it's off-chain. Too use Lightning, participants jointly lock funds on Bitcoin with an n-of-n multisignature. Then, they engage in a number of transactions between themselves. Each new "commitment transaction" splits those joint funds in a different way; these transactions are partially signed but they aren't put on the blockchain.
If you have a mass of unpublished transactions, any of which could be placed on the Blockchain, how do you keep one of the participants from reverting back to an old transaction that's more beneficial to them? The answer is revocation. A simplified example in BIP 112, which offers one of the stepping stones to Lightning, shows how. You give the participant who would be harmed by reversion to a revoked transaction the ability to reclaim the funds himself if the the other participant illegitimately tries to use the revoked transaction.
For example, presume that Alice and Bob update the commitment transaction to give more of the funds to Bob (effectively: Alice sent funds to Bob via this proto-Lightning network). They partially sign new transactions, but they also each offer up their own revokeCode
for previous transactions. This effectively guarantees that they won't publish previous transactions, because doing so would allow their counterparty to claim those previous funds.
So what does the old transaction look like? It was a commitment transaction showing funds intended for Alice, before she gave them to Bob. It had a locking script as follows:
OP_HASH160
<revokeHash>
OP_EQUAL
IF
<pubKeyBob>
ELSE
<+24Hours>
OP_CHECKSEQUENCEVERIFY
OP_DROP
<pubKeyAlice>
ENDIF
OP_CHECKSIG
The ELSE
block is where Alice got her funds, after a 24-hour delay. However now it's been superceded; that's the whole point of a Lightning-style payment channel, after all. In this situation, this transaction should never be published. Bob has no incentive to because he has a newer transaction, which benefits him more because he's been sent some of Alice's funds. Alice has no incentive either, because she loses the funds if she tries because of that revokeCode
. So no one puts the transaction onto the blockchain, and the off-chain transactions continue.
It's worth exploring how this script would work in a variety of situations, most of which involve Alice trying to cheat by reverting to this older transaction, which depicts the funds before Alice sent some of them to Bob.
Alice could try to use revocation code that she gave to Bob to immediately claim the funds. She writes a locking script of <sigAlice> <revokeCode>
:
Script: <sigAlice> <revokeCode> OP_HASH160 <revokeHash> OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Stack: [ ]
Script: OP_HASH160 <revokeHash> OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Stack: [ <sigAlice> <revokeCode> ]
Script: <revokeHash> OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Running: <revokeCode> OP_HASH160
Stack: [ <sigAlice> <revokeHash> ]
Script: OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Stack: [ <sigAlice> <revokeHash> <revokeHash> ]
Script: IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Running: <revokeHash> <revokeHash> OP_EQUAL
Stack: [ <sigAlice> True ]
The OP_EQUAL
feeds the IF
statement. Because Alice uses the revokeCode
, she gets into the branch that allows her to redeem the funds immediately, collapsing the rest of the script down to <pubKeyBob>
(within the conditional) and OP_CHECKSIG
(afterward).
Script: <pubKeyBob> OP_CHECKSIG
Running: True IF
Stack: [ <sigAlice> ]
Curses! Only Bob can sign immediately using the redeemCode
!
Script: OP_CHECKSIG
Stack: [ <sigAlice> <pubKeyBob> ]
Script:
Running: <sigAlice> <pubKeyBob> OP_CHECKSIG
Stack: [ False ]
So what if Alice instead tries to use her own signature, without the revokeCode
? She uses an unlocking script of <sigAlice> <notRevokeCode>
.
Script: <sigAlice> 0 OP_HASH160 <revokeHash> OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Stack: [ ]
Script: OP_HASH160 <revokeHash> OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Stack: [ <sigAlice> 0 ]
Script: <revokeHash> OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Running: 0 OP_HASH160
Stack: [ <sigAlice> <0Hash> ]
Script: OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Stack: [ <sigAlice> <0Hash> <revokeHash> ]
Script: IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Running: <0Hash> <revokeHash> OP_EQUAL
Stack: [ <sigAlice> False ]
We now collapse down to the ELSE
statement and what comes after the conditional:
Script: <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> OP_CHECKSIG
Running: False IF
Stack: [ <sigAlice> ]
Script: OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> OP_CHECKSIG
Stack: [ <sigAlice> <+24Hours> ]
And then Alice is foiled again because 24 hours haven't gone by!
Script: OP_DROP <pubKeyAlice> OP_CHECKSIG
Running: <+24Hours> OP_CHECKSEQUENCEVERIFY
Stack: [ <sigAlice> <+24Hours> ] — Script EXITS
What this means is that Bob has 24 hours to reclaim his funds if Alice ever tries to cheat, using the <revokeCode>
and his signature as his unlocking script:
Script: <SigBob> <revokeCode> OP_HASH160 <revokeHash> OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Stack: [ ]
Script: OP_HASH160 <revokeHash> OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Stack: [ <SigBob> <revokeCode> ]
Script: <revokeHash> OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Running: <revokeCode> OP_HASH160
Stack: [ <SigBob> <revokeHash> ]
Script: OP_EQUAL IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Stack: [ <SigBob> <revokeHash> <revokeHash> ]
Script: IF <pubKeyBob> ELSE <+24Hours> OP_CHECKSEQUENCEVERIFY OP_DROP <pubKeyAlice> ENDIF OP_CHECKSIG
Running: <revokeHash> <revokeHash> OP_EQUAL
Stack: [ <SigBob> True ]
Script: <pubKeyBob> OP_CHECKSIG
Running: True IF
Stack: [ <SigBob> ]
Script: OP_CHECKSIG
Stack: [ <SigBob> <pubKeyBob> ]
Script:
Running: <SigBob> <pubKeyBob> OP_CHECKSIG
Stack: [ True ]
All of Alice's commitment transactions are locked with this same locking script, whether they've been revoked or not. That means that the newest commitment transaction, which is the currently valid one, is locked with it as well. Alice has never sent a newer transaction to Bob and thus never sent him the previous revokeCode
.
In this situation, she could virtuously publish the transaction, closing down the proto-Lightning channel. She puts the transaction on the chain and she waits 24 hours. Bob can't do anything about it because he doesn't have the recovation code. Then, after the wait, Alice reclaims her funds. (Bob does the same thing with his own final commtiment transaction.)
The Revocable Commitment Transactions were just a stepping stone to Lightning. The actual Lightning Network uses a more complex mechanism called a hashed timelock contract, or HTLC.
The main purpose of HTLCs is to create a comprehensive network of participants. Transactions are no longer just between a pair of participants who have entered the network together, but can now be between previously unassociated people. When funds are sent, a string of transactions are created, each of them locked with a secretHash
. When the corresponding secretCode
is revealed, the entire string of transactions can be spent. This is what allows singular transactions to actually become a network.
There's also a bit more complexity in Lightning Network locking scripts. There are separate locks for the sender and the recipient of each transaction that are more widely divergent than the differing commitment transactions alluded to in the previous section. We're going to show both of them, to demonstrate the power of these locking scripts, but we're not going to dwell on how they interact with each other.
Once more, we're going to start looking at Alice's commitment transaction, which shows funds that she's received:
OP_HASH160
OP_DUP
<secretHash>
OP_EQUAL
IF
<+24Hours>
OP_CHECKSEQUENCEVERIFY
OP_2DROP
<pubKeyAlice>
ELSE
<revokeHash>
OP_EQUAL
OP_NOTIF
<Date>
OP_CHECKLOCKTIMEVERIFY
OP_DROP
ENDIF
<pubKeyBob>
ENDIF
OP_CHECKSIG
The key to these new HTLCs is the secretHash
, which we've seen is what allows a transaction to span the network. When the transaction has spanned from its originator to its intended recipient, the secretCode
is revealed, which allows all the participants to create a secretHash
and unlock the whole network of payments.
After the secretCode
has been revealed, the IF
branch opens up: Alice can claim the funds 24 hours after the transaction is put on the Bitcoin network.
However, there's also the opportunity for Bob to reclaim his funds, which appears in the ELSE
branch. He can do so if the transaction has been revoked (but Alice puts it on the blockchain anyway), or if an absolute timeout has occurred.
Here's the alternative commitment transaction locking script used by the sender:
OP_HASH160
OP_DUP
<secretHash>
OP_EQUAL
OP_SWAP
<revokeHash>
OP_EQUAL
OP_ADD
IF
<pubKeyAlice>
ELSE
<Date>
OP_CHECKLOCKTIMEVERIFY
<+24Hours>
OP_CHECKSEQUENCEVERIFY
OP_2DROP
<pubKeyBob>
ENDIF
OP_CHECKSIG
The initial part of their Script is quite clever and so worth running:
Initial Script: <suppliedCode> OP_HASH160 OP_DUP <secretHash> OP_EQUAL OP_SWAP <revokeHash> OP_EQUAL OP_ADD
Stack: [ ]
Initial Script: OP_HASH160 OP_DUP <secretHash> OP_EQUAL OP_SWAP <revokeHash> OP_EQUAL OP_ADD
Stack: [ <suppliedCode> ]
Initial Script: OP_DUP <secretHash> OP_EQUAL OP_SWAP <revokeHash> OP_EQUAL OP_ADD
Running: <suppliedCode> OP_HASH160
Stack: [ <suppliedHash> ]
Initial Script: <secretHash> OP_EQUAL OP_SWAP <revokeHash> OP_EQUAL OP_ADD
Running: <suppliedHash> OP_DUP
Stack: [ <suppliedHash> <suppliedHash> ]
Initial Script: OP_EQUAL OP_SWAP <revokeHash> OP_EQUAL OP_ADD
Stack: [ <suppliedHash> <suppliedHash> <secretHash> ]
Initial Script: OP_SWAP <revokeHash> OP_EQUAL OP_ADD
Running: <suppliedHash> <secretHash> OP_EQUAL
Stack: [ <suppliedHash> <wasItSecretHash?> ]
Initial Script: <revokeHash> OP_EQUAL OP_ADD
Running: <suppliedHash> <wasItSecretHash?> OP_SWAP
Stack: [ <wasItSecretHash?> <suppliedHash> ]
Initial Script: OP_EQUAL OP_ADD
Stack: [ <wasItSecretHash?> <suppliedHash> <revokeHash> ]
Initial Script: OP_ADD
Running: <suppliedHash> <revokeHash> OP_EQUAL
Stack: [ <wasItSecretHash?> <wasItRevokeHash?> ]
Initial Script:
Running: <wasItSecretHash?> <wasItRevokeHash?> OP_ADD
Stack: [ <wasItSecretOrRevokeHash?> ]
Running through the script reveals that the initial checks, above the IF
/ELSE
/ENDIF
determine if the hash was either the secretCode
or the revokeCode
. If so, Alice can take the funds in the first block. If not, Bob can take the funds, but only after Alice has had her chance, and both the 24 hour timeout and the absolute timeout have passed.
HTLCs are quite complex, and this overview doesn't try to explain all of their intricacies. Rusty Russell's overview explains more. and there's even more detail in his Deployable Lightning paper. But don't worry if some of the intricacies still escape you, particularly the interrelations of the two scripts.
For the purposes of this tutorial, there are two important lessons for HTLCs:
- Understand that a very complex structure like an HTLC can be created with Bitcoin Script.
- Analyze how to run each of the HTLC scripts.
It's worth your time running each of the two HTLC scripts through each of its permutations, one stack item at a time.
We're closing our examination of Bitcoin Scripts with a look at how truly powerful they can be. In 20 opcodes or less, a Bitcoin Script can form the basis of an entire off-chain payment channel. Similarly, two-way pegged sidechains are the product of less than twenty opcodes, as also briefly noted in BIP 112.
If you've ever seen complex Bitcoin functionality or Bitcoin-adjacent systems, they were probably built on Bitcoin Scripts. And now you have all the tools to do the same yourself.
Move on to "Bitcoin APIs" with Chapter Twelve: Talking to Bitcoind.